Epidemiology of eltor cholera in rural Bangladesh: importance of surface water in transmission

BUletin of the World Health Organization, 60 (3): 395- 404 (1982) Epidemiology of eltor cholera in rural Bangladesh: importance of surface water in transmission JAMES M. HUGHES,' JOHN M. BOYCE,2 RICHARD J. LEVINE,3 MOSLEMUDDIN KHAN, K. M. A. Aziz,5 M. I. HUQ,4 & GEORGE T. CURLIN6 In order to define the role ofwater usedfor drinking, cooking, bathing, and washing in the transmission ofVibrio cholerae biotype eltor infections in an area with endemic cholera, surveillance was initiated in neighbourhoods with a culture-confirmed cholera index case and others with index cases with non-cholera diarrhoea as controls. In neighbourhoods with cholera infection, 44% ofsurface water sources werepositivefor V. cholerae, whereas only 2% ofsurface sources were positive in control neighbourhoods. Canals, rivers, and tanks were most frequently positive. There was an increased risk of infection for families using water from culture-positive sources for drinking, cooking, bathing, or washing and for those using water sources used by indexfamiliesfor drinking, cooking or bathing. Analysis ofthe resultsfor individuals showed that in this case there was an increased risk ofinfection associated with using waterfrom culture-positive sourcesfor cooking, bathing, or washing, but not with using waterfrom culture-positive sourcesfor drinking. Individuals who used the same water source as an indexfamily for bathing were more likely to be infected than those using different sources. For families drinking from a culture-negative source, there was an association between infection and bathing in a positive source. Forfamilies using a different bathing sourcefrom the indexfamily there was an association between infection and drinking from the same source as the indexfamily, andforfamilies using a different drinking source from the index family there was an association between infection and bathing in thesame source as the indexfamily. These data suggest that use ofsurface water is important in the transmission ofV. cholerae and that, in addition toproviding safe drinking water, education regarding the risk of transmission of infection by waterfrom potentially contaminated sources usedfor otherpurposes, especially bathing, may also be necessary to control transmission in areas where eltor cholera is endemic. Transmission of Vibrio cholerae by contaminated drinking water was first documented 130 years ago (1). However, the importance of drinking water in transmission of cholera in an area with endemic cholera in rural Bangladesh is still debated (2, 3) because the anticipated reduction in cholera case rates has not been observed for persons with easy access to ' Assistant Chief, Enteric Diseases Branch, Bacterial Diseases Division, Center for Infectious Diseases, Centers for Disease Control (CDC), Atlanta, GA 30333, USA. 2 Medical Epidemiologist, Field Services Division, Bureau of Epidemiology, CDC, Atlanta. Present address: Assistant Professor, Division of Infectious Diseases, Department of Medicine, University of Mississippi Medical Center, Jackson, MS 39216, USA. 3 Medical Epidemiologist, Epidemiology Division, Cholera Research Laboratory (now ICDDR,B), Dacca, Bangladesh. Present address: Chief, Department of Epidemiology, Chemical Industry Institute of Toxicology, Research Triangle Park, NC 27709, USA. 4 Scientist, International Centre for Diarrhoeal Disease Research (ICDDR,B), Dacca, Bangladesh. 5 Associate Scientist, ICDDR,B. 6 Chief, Epidemiology Division, Cholera Research Laboratory (now ICDDR,B). Present address: Chief, Health, Population, and Nutrition Division, Office of Technical Resources, Asia Bureau, United States Agency for International Development, Washington, DC 20523, USA. tube-well water of presumed good microbiological quality (4) or for those who usually drink tube-well water (5- 7). A number of hypotheses have been proposed to explain the failure of the use of tube-well water to protect individuals from cholera infection (8), but none have satisfactorily explained the relationship of water use to cholera transmission. To evaluate the relation between water used for drinking and for other purposes and the transmission of cholera in Matlab Bazar Thana, a rural area in Bangladesh (9), longitudinal studies were conducted in neighbourhoods with a resident who had symp- tomatic, culture-confirmed V. cholerae infec- tion and in control neighbourhoods with a resident with non-cholera diarrhoea. The methods used varied with community size. Studies in small neighbour- hoods were designed to assess the role of water in cholera transmission and to define the clinical spec- trum of cholera. Studies in large neighbourhoods were designed to assess transmission of V. cholerae in families of persons with cholera. The results reported in this paper confirm the observation that surface 4189 -395 J. M. HUGHES ET AL. water in proximity to persons with V. cholerae infec- tion is frequently contaminated with V. cholerae (10, 11) and suggest that the water used for drinking and other purposes, especially bathing, is important in cholera transmission in this environment. MATERIALS AND METHODS Index patients were selected from persons who had diarrhoea, lived in villages in the Matlab vaccine trial surveillance (VTS) area, and were seen as either in- patients or outpatients at the Cholera Hospital. Each candidate for the study had a positive rectal swab culture for V. cholerae the day the neighbourhood study was begun; if more than one patient had a positive culture, we used a table of random numbers to select the index case. Patients' residences were visited the morning after presentation at the hospital, frequently before the patient had returned home. Using a table of random numbers, we selected control patients from all inpatients and outpatients with diar- rhoea who lived in the VTS area and had a negative rectal swab culture for V. cholerae the day each study was begun. Study areas included all families living in a bari (a cluster of houses located on the same elevated court- yard or compound and inhabited by patrilineally related families) in which the index patient lived. A neighbourhood was defined as all households on the same elevated piece of land and was arbitrarily classified as "small" if it contained 17 or fewer fam- ilies and "large" if it contained 18 or more families. Studies were conducted in 14 cholera and 14 control neighbourhoods between December 1973 and Febru- ary 1974 during the cool dry post-monsoon period. Ten cholera and 9 control neighbourhoods were defined as small. A family was defined as all individuals sharing the same dwelling or dwellings and eating food prepared in one kitchen. After obtaining the free and informed consent of all families in each neighbourhood, a questionnaire was completed on the first day of the study by interviewing an adult family member, usually a female. Information was obtained on the age and sex of all family members and guests and the source(s) of water currently used by all persons in the family for drinking, cooking, bathing, and washing clothes and utensils. Tanks were defined as rectangular ponds formed when monsoon rains fill depressions resulting from removal of earth to build mounds on which houses were built. Tanks retain water all the year round. Ditches were defined as smaller depressions that contain water during part of the year. Information about stool patterns of each indi- vidual during the previous week was obtained on the first day; on subsequent days, information was obtained about stool patterns during the previous 24 hours, either by interviewing the individuals con- cerned or an adult family member. Diarrhoea was defined as a history of stools that were looser or more watery or occurred more frequently than normal for that individual. Cholera infection was diagnosed on the basis of a positive culture for V. cholerae. A clinical case of cholera was defined as the presence of diarrhoea within one day of the person's having a positive culture for V. cholerae. In small neighbourhoods, rectal swabs were obtained each day of the study from all available residents; when a positive culture was obtained, no more swabs were taken from that person. If an indi- vidual had a second epidsode of diarrhoea (defined as being separated by at least two days with a normal stool pattern), daily culturing was resumed. In large neighbourhoods rectal swabs were taken on three con- secutive days from all individuals with diarrhoea. Rectal swabs were also obtained each day from all family contacts of culture-positive individuals throughout the remainder of the study. Studies were discontinued when no new culture-positive indi- viduals were identified for nine consecutive days. Data for individuals were included in the question- naire survey if they were questioned on at least five days or had a history of diarrhoea. They were included in the culture survey if they had a positive culture or three negative cultures for V. cholerae. Rectal swabs were cultured for V. cholerae, and iso- lates were biotyped as described by Sommer & Wood- ward (4). Approximately 60 ml of water was collected from each tube well, ditch, tank, canal, and river used by residents of both small and large neighbourhoods for drinking, cooking, bathing, or washing. Indi- vidual samples were obtained on several days begin- ning on the first day of each study near the periphery of each source at a site where villagers obtained water. Each water sample was placed in approximately 30 ml of triple-strength bile peptone water; tellurite was added to a dilution of 1:200 000, and the sample was incubated at 37 °C for 6 hours and then subcultured onto tellurite-taurocholate-gelatin (TTGA) agar. Colonies thought to be V. cholerae were identified (4). Cultures were recorded as either positive or negative; no quantitative results were obtained. Sources were considered adequately cultured if a positive culture or at least three negative cultures were obtained. Statistical analyses were performed using the two- tailed Fisher's exact test, the chi-square test with Yates' correction, the t-test on proportions with variance adjusted for cluster effect, and the Mantel- Haenszel test. 396 ROLE OF SURFACE WATER IN CHOLERA TRANSMISSION RESULTS In the small cholera neighbourhoods, 504 of 617 individuals (82%) participated in the questionnaire survey and 476 of 617 (77/o) participated in the culture survey. When the 10 index cases are eliminated, only one of 56 infections (2%) identified in small neighbourhoods involved diarrhoea severe enough to require medical attention whereas 32 (57%) involved mild diarrhoea, and 23 (41%) were asymptomatic. All infections were caused by V. cholerae biotype eltor. Cholera infection rates were similar for all age groups. Case rates were highest in individuals 5 - 9 years of age and were significantly higher than the rate in adults (Table 1). Sex-specific rates for sympto- matic or asymptomatic infection were comparable. Ninety-four percent of infections and 98% of cases were identified by day 12 of the studies (Fig. 1). By day five, only 59% of infections and 597o of cases had been identified. Rates of infection identified in both small and large neighbourhoods were significantly higher for index families (23%) than other families (8%) in the neigh- bourhood (P < 0.001); rates were higher in the index bari (11 ), when data for the index family were excluded, than in other baris in the neighbourhood (4/o) (P < 0.01). Similar studies conducted in 14 control neighbourhoods did not document a single V. cholerae infection. Eleven of 14 (79%) neighbourhoods with an index V. cholerae infection had at least one water source contaminated with V. cholerae, as did only one of 14 (7%) control neighbourhoods (P < 0.001). In cholera-infected neighbourhoods, 44% of all cultured surface water sources were positive for V. cholerae, as were only 2% of surface sources in Table 1. Infection and case rates, by age group, in small neighbourhoods Age group No. No. infected No. ill (years) cultured (%) M%) <1 30 4(13.3) 2(6.7) 1-4 76 11 (14.5) 6(7.9) 5-9 79 12 (15.2) 10 (12.7)a 10-14 76 10 (13.2) 6 (7.9) > 15 182 19(10.4) 9(4.9)a Total 443b 56 (12.6) 33 (7.4) a P= 0.037 b Index cases excluded. M AsYMPOM"ATC 4 Co X:- _.. EI2345673 *I0II23I45167I6320u2l jTi s W s | r ' @139' 10a8'4'Isr'lVIVS .0212e DAY OF STUDY Fig. 1. Vibrio cholerae cases and asymptomatic infec- tions, by day of detection, large and small neighbour- hoods, Bangladesh. control neighbourhoods (Table 2). Rivers were most frequently positive, and then canals, tanks, and ditches in that order. Positive cultures for V. cholerae were obtained from one tube well on days 2, 3, and 4 of one study. Cultures of water obtained from this well on day 1 and on days 5- 10 were negative. Cul- tuI;es of water obtained from two tube wells in two other cholera neighbourhoods were negative. Similar types of water source were available in cholera and control neighbourhoods. When the percentages of families in cholera and control neighbourhoods that used water from each of the five types of source were compared (Table 3), more families in neighbourhoods with a cholera index case used canal and river water for drinking, cooking, bathing, and washing, or water from ditches for washing. Conversely, significantly fewer families in neighbourhoods with a cholera index case than families in control neighbourhoods used water from tanks for cooking, bathing, and washing, and water from tube wells for drinking. When the choices of water source for drinking, cooking, bathing, and washing in small neighbour- hoods with a cholera index case were evaluated, only one significant association was found. Families with an infected individual were more likely to drink from tube wells than were families without infection (P < 0.01); however, only eight infected families used tube-well water for drinking. Six of these eight infected families lived in the same neighbourhood, which had a contaminated tube well, and the families drank water from that well. In small neighbourhoods with a cholera index case, families using at least one culture-positive water source for drinking, cooking, bathing, or washing were significantly more likely to have a person 397 J. M. HUGHES ET AL. Table 2. Results of water cultures for V. cholerae in small and large neighbourhoods, with and without cholera infection Cholera neighbourhoods (n = 14) Control neighbourhoods (n = 14) No. Total No. Total with no. of No. No. % with no. of No. No. % Water source source sources cultured positive positive source sources cultured positive positive Surface water River 5 5 4 3 75 7 7 4 0 0 Canal 8 8 8 5 63 7 7 5 0 0 Tank 9 23 22 10 44a 1 1 44 41 1 2a Ditch 7 21 14 3 21 6 9 6 0 0 Subtotal - 57 48 21 44 - 67 56 1 2 Tubewell 4 4 3 1 33 6 8 7 0 0 a p < 0.001. Table 3. Percentage of families using water sources, by type and purpose, in small and large neighbourhoods, with and without cholera infection Tubewell Ditch Tank Canal River Drinking Cholera 12a 5 16 26a 53a Control 34 8 21 12 36 Cooking Cholera 0 21 41 5i° 280 Control 0 14 770 1 1 10 Bathing Cholera 0 7 34 28° 510 Control 0 13 610° 13 33 Washing Cholera 0 31 43 320 23° Control 0 150 83° 11 6 0 Significantly different from control, P < 0.001. infected with V. cholerae than those using only culture-negative sources (Table 4). Infection rates for families using culture-positive sources for drinking and other purposes (51 No) and those for families using culture-negative sources for drinking but culture- positive sources for cooking, bathing, or washing (58%) were similar; however, infection rates for both groups were higher than for families using culture- negative sources for all purposes (14%, P < 0.001, and P < 0.01, respectively). Families using the same water source(s) as the index family for drinking, cook- ing, or bathing were significantly more likely to have a person infected with V. cholerae than those using different sources (Table 5). Analysis of data for individuals in small neighbour- hoods with a cholera index case revealed that persons who drank from at least one culture-positive source were not significantly more likely to be infected than those who drank only from negative sources. However, those who cooked with, bathed in, or washed with water from at least one positive source were significantly more likely to be infected than those who used only negative water sources for these purposes (Table 6). The proportion of individuals using water from culture-positive sources for drink- ing and other users who were infected was similar (0.167) to that for those using water from culture- negative sources for drinking, but culture-positive sources for cooking, bathing, and/or washing (0.194). Individuals who used the same water sources as those in the index family for bathing were signifi- cantly more likely to be infected than those who used different sources (Table 7). Individuals who used the same sources as members of the index family for drinking, cooking, or washing were no more likely to be infected than those who used different sources. In an attempt to define the relative importance of drinking and bathing in the transmission of V. cholerae, families were stratified according to positivity of the drinking and bathing source (Table 8) and according to whether they used the same sources as thejndex family for drinking or bathing (Table 9). For families bathing in positive sources (Table 8), there was no association between infection and use of 398 ROLE OF SURFACE WATER IN CHOLERA TRANSMISSION Table 4. Family infection rates, according to culture status of water sources, small neighbourhoods Total no. of No. of families Percentage P Water use families infected infected value Drank water from > 1 positive source 43 22 51.2 0.017 Drank water from negative sources 47 12 25.5 Cooked with water from > 1 positive source 54 29 53.7 0.0001 Cooked with water from negative sources 36 5 13.9 Bathed in > 1 positive source 55 29 52.7 0.0003 Bathed in negative sources 35 5 14.3 Washed with water from > 1 positive source 55 29 52.7 0.0003 Washed with water from negative sources 35 5 14.3 Table 5. Family infection rates, according to use of index-family water source, small neighbourhoods Total no. of No. Percentage P Water use families infected infected value Drank from same sources 55 22 40.0 0.004 Drank from different sources 25 2 4.0 Cooked from same sources 41 17 41.5 0.03 Cooked from different sources 39 7 17.9 Bathed in same sources 50 19 38.0 0.05 Bathed in different sources 30 5 16.7 Washed in same sources 41 15 36.6 0.09 Washed in different sources 39 9 23.1 Table 6. Proportion of individuals infected, according to culture status of water source, small neighbourhoods Total no. of No. Proportion infected Water use individuals infected (SE) Drank water from > 1 positive source 265 44 0.166 (0.027) Drank water from negative sources 211 22 0.104 (0.032) Cooked with water from > 1 positive source 328 57 0.174 (0.026)ff Cooked with water from negative sources 148 9 0.061 (0.029) Bathed in > 1 positive source 332 57 0.172 (0.026)a Bathed in negative sources 144 9 0.063 (0.030) Washed with water from > 1 positive source 332 57 0.172 (0.026) Washed with water from negative sources 144 9 0.063 (0.030) a ' Significantly different from those using negative sources, P < 0.01. 399 J. M. HUGHES ET AL. Table 7. Proportion of individuals infected, according to use of water source used by index family, small neighbour- hoods Total no. of No. Proportion infected Water use individuals infected (SE) Drank water from same sources 295 37 0.125 (0.022) Drank water from different sources 127 7 0.055 (0.034) Cooked with water from same sources 223 28 0.126 (0.024) Cooked with water from different sources 199 16 0.080 (0.028) Bathed in same sources 255 34 0.133 (0.026) a Bathed in different sources 167 10 0.060 (0.024) Washed with water from same sources 207 25 0.121 (0.025) Washed with water from different sources 215 19 0.088 (0.027) a Significantly different from those using different sources, P < 0.05. Table 8. Family infection rates, according to culture positivity of drinking and bathing water sources, small neighbourhoods Drinking Bathing Total no. No. Percentage source source families infected infected + + 43 22 51.20 + - 0 0 - - + 12 7 58.3 - - 35 5 14.3b a For families with culture-positive bathing source, not significant. b For families with culture-negative drinking source, P = 0.005. positive sources for drinking. In contrast, for families drinking from negative sources, there was an associ- ation between infection and use of positive sources for bathing. For families using the same source as the index family for bathing (Table 9), there was no association between infection and drinking from the same source used by the index family; however, for those bathing in different sources, there was an association between infection and using the same source used by the index family for drinking. For families drinking from the same source as the index family, there was no association between infection and bathing in the same source used by the index family; however, for those drinking from sources different from the index family, there was an associ- ation between infection and bathing in the same source used by the index family. For all families, Table 9. Family infection rates, according to use of same or different water sources used by index family for drink- ing and bathing, small neighbourhoods Drinking Bathing Total no. No. Percentage source source families infected infected same same 46 17 37.Oa c same different 9 5 55.6b c different same 4 2 50.Oa d different different 21 0 ob d Mantel-Haenszel test: P = 0.06 drinking (controlling for bathing) P = 0.88 bathing (controlling for drinking) a For families with same bathing source as index family, not significant. b For families with bathing source different from index family, P= 0.001. For families with same drinking source as index family, not significant. d For families with drinking source different from index family, P = 0.02. Mantel-Haenszel analysis indicated that there was a borderline association between drinking and infec- tion when controlling for bathing but not between bathing and infection when controlling for drinking. DISCUSSION Previously reported studies of the epidemiology of V. cholerae biotype eltor infections in Matlab Bazar have failed to demonstrate convincingly a significant 400 ROLE OF SURFACE WATER IN CHOLERA TRANSMISSION role for contaminated drinking water in transmission (4- 7). Explanations include the possibility that cholera is not primarily water-borne in this environment, that people report using tube-well water when they do not, that only some family members use tube-well water while others use alternative sources, or that the protection provided by use of tube-well water is negated by the use of contaminated surface water for cooking, bathing, or washing (7, 8). Results of this study, although qualified by the possibility that not all water sources used by each family member were specified and by the lack of information on water storage practices and possible contamination in the home, suggest that contaminated surface water used for drinking and for other purposes, especially bathing, is important in transmission of V. cholerae in this environment. Rates of infection with V. cholerae in the neigh- bourhood studies were comparable for all age groups. From 1968 - 77, the age-specific cholera case rate for patients seen at the Matlab Hospital was highest for children 2-4 years old (4.5/1000) followed by that for children 5 - 9 years old (3.8/1000) (12). Results of our neighbourhood surveillance indicate that case rates were highest for children 5 -9 years old and were significantly higher for this group than for persons aged 15 years or over but not significantly higher than for other age groups. These differences probably reflect the different case detection tech- niques in the two studies. In our study, no V. cholerae infections were detected in control neighbourhoods, an observation consistent with earlier reports that classical V. cholerae cases occurred in localized outbreaks in both Dacca (13) and Matlab (9). The highest infection rate was in cholera index families, in which 23%o of individuals were infected. This infection rate is comparable with infection rates of 11%o, 13%o, 21 7o, and 25%o reported for families with classical V. cholerae infection (9, 14- 16) and with infection rates of 21% and 31% reported for families with V. cholerae biotype eltor infection in Bangladesh (16, 17). However, in contrast to obser- vations in the study of eltor infection in Dacca, which indicated that none of 136 neighbourhood residents was infected, index bari residents were significantly more likely ( 107) than residents of other baris in the same neighbourhood (4%) to be infected. Only 59% of all infections were identified by day 5 of the study in each bari, in contrast with 86% of all infections during the first five days of a ten-day study of classical cholera infections in Dacca (16). These observations suggest that V. cholerae biotype eltor frequently infects other members of index households and baris in this environment and that the risk of acquiring infection exists for at least 1 - 2 weeks. It is well documented that eltor strains survive longer than classical strains in both water and food (18). In one study, V. cholerae biotype eltor survived 19.3 ± 5.1 days in shallow well-water whereas classical strains survived 7.5 ± 1.9 days (18). The isolation of eltor strains from 44% of surface water sources in infected neighbourhoods is comparable with the 45% rate reported from an urban environ- ment where classical V. cholerae cases were occurring (13) and with the 57% rate in Matlab in neighbour- hoods with infections caused by eltor organisms (11). The observation that over 60/o of canals and rivers used by cholera-infected neighbourhoods were positive is noteworthy because organisms entering these bodies of water would be diluted; however, this observation is consistent with earlier observations that use of water from these sources is associated with increased risk of classical (10) and V. cholerae biotype eltor infections (5). Although the availability of canals and rivers in neighbourhoods with and without cholera index cases was comparable, families in neighbourhoods with cholera were significantly more likely to use canal and river water for drinking as well as for cooking, bathing, and washing. Whether water usage patterns might reflect differences in socio- economic status which might in turn influence suscep- tibility to V. cholerae infection was not addressed. Only one water source in a neighbourhood not infected with V. cholerae was positive; in a study conducted in Matlab in 1965 -66, only 10 of 5670 surface water cultures were positive for classical V. cholerae, and all positive samples were from neigh- bourhoods with documented V. cholerae infection (9). One of three tube wells tested by culture was positive for V. cholerae on three consecutive days. The use of water from this well by six families that had at least one person infected with V. cholerae accounts for the observed association of tube-well use with infection. This tube well was in good repair and was not primed by adding water. However, neighbour- hood residents admitted to hanging their laundry on the pump to dry. Contaminated water may have leaked into the well around the vertical pump rod. The observation of Spira et al. that none of 12 wells they cultured was contaminated (11) suggests that tube-well water contamination with V. cholerae is uncommon. The importance of water in the transmission of cholera is illustrated by the fact that, in small neigh- bourhoods where case-finding techniques were most intensive, families who used a culture-positive water source for drinking, cooking, bathing, or washing were signficantly more likely to have a cholera infection than families who used negative sources. These findings are compatible with observations by Levine et al. that high cholera case rates were associated with use of contaminated canal water (19) and by Spira et al. that the infection rate for families J. M. HUGHES ET AL. increased as the percentage of positive water cultures in the home increased (11). Spira's observations strongly suggested that water was contaminated almost exclusively at the source rather than in the home. We found that families who used the same water source(s) as the index family for either drink- ing, cooking or bathing were more likely to have an infected individual. Several findings imply that drinking contaminated water may not be the only important factor in the transmission of eltor cholera, a hypothesis suggested previously by others (4, 7, 10, 11). Families who drank from culture-positive sources were no more likely to be infected that those who drank from culture-negative sources but used positive sources for cooking, bathing, or washing; both groups were more likely to be infected than families using only culture- negative sources. In addition, when data from indi- viduals rather than families were considered, those who drank from culture-positive sources were no more likely to be infected than those who drank from culture-negative ones but used culture-positive sources for other purposes. Individuals who used the same water source as members of the index family for bathing were significantly more likely than those using different sources to be infected. For families with culture-negative drinking sources, the association between infection and bathing in a positive source suggests that bathing in contaminated water is a risk factor for infection, independent of drinking, but could also be inter- preted as indicating that infected persons defaecate where they bathe. However, the association between infection and drinking from the same source as the index family, in families bathing in different sources from the index family, and between infection and bathing in the same source as the index family, in families drinking from different sources than the index family, suggest that both drinking and bathing in contaminated water are risk factors for V. cholerae infection in this environment. The association of bathing and infection is compatible with a previous suggestion that ingestion of water while bathing may be important in the transmission of classical cholera (10) and raises a question concerning the relative concentrations of V. cholerae in contaminated water stored in the home and that at bathing sites. This hypothesis is compatible with the custom in West Bengal of taking handfuls of water into the mouth during bathing (20). In summary, extensive environmental contami- nation occurred in neighbourhoods with residents who were infected with V. cholerae. Although not providing information regarding the mode of intro- duction of V. cholerae into a neighbourhood, the data support a role for contaminated water in transmission once the organism is introduced. These observations suggest that, in addition to providing safe drinking water, the education of persons about the risk of transmission of infection by water from potentially contaminated sources used for other purposes, especially bathing, may be necessary to control transmission in areas where V. cholerae biotype eltor is endemic. RESUME tPIDEMIOLOGIE DU CHOLERA ELTOR DANS LES ZONES RURALES DU BANGLADESH: LE ROLE DE L'EAU DE SURFACE DANS SA TRANSMISSION La transmission de Vibrio cholerae par de l'eau de boisson contamin6e a ete demontr6e pour la premiere fois il y a 130 ans. Le r8le de l'eau de boisson dans la transmission du chol6ra dans une r6gion end6mique rurale du Bangladesh est toutefois encore controverse. Cette 6tude devait per- mettre de d6terminer le r8le de l'eau utilis6e pour la boisson, la cuisine, les ablutions et la lessive dans la transmission de V. cholerae biotype eltor dans la r6gion rurale de Matlab Bazar Thana au Bangladesh. On a proc6d6 a des 6tudes longitudinales dans des zones ou des cas signaux de chol6ra avaient ete d6tect6s et confirm6s par culture, et dans des zones t6moins oti les cas signaux de diarrh6e n'etaient pas li6s au chol6ra. Au cours de visites quotidiennes chez les habitants de ces deux zones, on a recueilli des ant6cedents sur les cas de diarrh&e, des cultures d'6couvillonnages rectaux et des 6chantillons d'eau utilis6e par les habitants pour des cultures de V. cholerae. Dans les zones o'u avaient ete enregistr6s des cas signaux de chol6ra, 44% des sources d'eau de surface 6taient conta- min6es par V. cholerae, alors que dans les zones t6moins ce pourcentage n'6tait que de 2%. Les canaux, les cours d'eau et les r6servoirs 6taient tres souvent contamin6s. Bien que l'on trouve les memes types de sources d'eau dans les r6gions de chol6ra et dans les r6gions t6moins, les families des r6gions touch6es avaient davantage tendance a utiliser l'eau des canaux et des cours d'eau pour la boisson, la cuisine, les ablutions et la lessive que les familles des zones t6moins. Les risques d'infection 6taient plus importants chez les familles utilisant de l'eau contamin6e pour la boisson, la cuisine, les ablutions et la lessive. Les taux d'infection 6taient toutefois les memes pour les families utilisant de l'eau contamin6e pour la boisson et pour d'autres usages et pour les familles qui consommaient de l'eau non contamin6e pour la boisson mais de l'eau contamin6e pour la cuisine, les 402 ROLE OF SURFACE WATER IN CHOLERA TRANSMISSION 403 ablutions et la lessive; dans ces deux derniers groupes, les taux d'infection 6taient plus 6lev6s que chez les familles n'utilisant que de l'eau non contamin6e. Le risque plus 6lev6 d'infection chez les families 6tait 6galement ie A la consom- mation pour la boisson, la cuisine ou les ablutions d'eau provenant de sources 6galement utilis6es par les familles indicatrices. Chez les individus, les risques plus 6lev6s d'infection etaient lies A la consommation d'eau provenant de sources contamin6es pour la cuisine, les ablutions ou la lessive, mais pas A la consommation d'eau de boisson contamin6e. Les individus qui utilisaient pour les ablutions les memes sources d'eau qu'une famille indicatrice 6taient plus expos6s que ceux utilisant de l'eau provenant d'autres sources. On a pu etablir chez les familles consommant de l'eau de boisson provenant d'une source non contamin6e un lien entre l'infection et l'utilisation d'eau contamin6e pour les ablutions. En ce qui concerne les familles n'utilisant pas pour leurs ablutions la meme source que la famille indica- trice, on a pu etablir un lien entre l'infection et ia consom- mation de la meme eau de boisson que la famille indicatrice; on a 6galement trouv6 pour les familles ne consommant pas la meme eau de boisson que la famille indicatrice une corr6lation entre l'infection et l'utilisation de la meime source pour les ablutions. Ces donn6es donnent i penser que l'eau de surface joue un role important dans la transmission de V. cholerae et que, outre le fait de permettre de disposer d'une eau de boisson saine, I'education en ce qui concerne les risques de transmission de l'infection par de l'eau provenant de sources potentiellement contamin6es et utilis6es a d'autres fins, notamment pour les ablutions, peut egalement etre n6cessaire pour lutter contre la transmission dans les zones oiu le cholera eltor est endemique. ACKNOWLEDGEMENTS We are grateful to Mr J. Chakraborty and Mr M. R. Khan who supervised the field staff during this study, to Mr A. R. M. A. Alim who assisted in identification of the isolates, to Dr Roger A. Feldman and Dr Michael H. Merson for their helpful suggestions during preparation of the manuscript, to Mr Robert A. Pollard for statistical assistance, and to the National Institutes of Health and the International Centre for Diarrhoeal Disease Research, Bangladesh, for support during the study. Finally, we are grateful to Ms Peggy Hutton and Ms Ruth Greenberg for assistance in preparing the manuscript. REFERENCES 1. SNOW, J. Snow on cholera. London, University Press, 1936. 2. FEACHEM, R. Is cholera primarily water-borne? Lancet, 2: 957 (1976). 3. LEVINE, R. & NALIN, D. Cholera is primarily water- borne in Bangladesh. Lancet, 2: 1305 (1976). 4. SOMMER, A. & WOODWARD, W. The influence of pro- tected water supplies on the spread of classical/Inaba and El Tor/Ogawa cholera in rural East Bengal. Lancet, 2: 985 - 987 (1972). 5. KHAN, M. ET AL. Water sources and the incidence of cholera in rural Bangladesh. Dacca, International Centre for Diarrhoeal Disease Research, 1978 (Scientific Report No. 16). 6. CURLIN, G. ET AL. The influence of drinking tubewell water on diarrhea rates in Matlab Thana, Bangladesh. Dacca, International Centre for Diarrhoeal Disease Research, Bangladesh, 1977 (Working Paper No. 1). 7. LEVINE, R. ET AL. Failure of sanitary wells to protect against cholera and other diarrhoeas in Bangladesh. Lancet, 2: 86- 89 (1976). 8. BRISCOE, J. The role of water supply in improving health in poor countries (with special reference to Bangladesh). Americanjournal ofclinical nutrition, 31: 2100-2113 (1978). 9. MCCORMACK, W. ET AL. Endemic cholera in rural East Pakistan. American journal of epidemiology, 89: 393-404 (1969). 10. KHAN, M. & MOSELEY, W. The role of boatman in the transmission of cholera. East Pakistan medicaljournal, 11: 61-65 (1967). 11. SpIRA, W. M. ET AL. Microbiological surveillance of intra-neighbourhood El Tor cholera transmission in rural Bangladesh. Bulletin of the World Health Organization, 58: 731 - 740 (1980). 12. MERSON, M. ETAL. Epidemiology of cholera and entero- toxigenic Escherichia coli diarrhea. In: Proceedings of the 43rd Nobel Symposium on Cholera and Related Diarrheas: Molecular Aspects of a Global Health Problem, Stockholm. 6-11 August 1978. Basel, S. Karger, 1980, pp. 34 - 45. 13. MARTIN, A. ET AL. Epidemiologic analysis of endemic cholera in urban East Pakistan, 1964- 1966. American journal ofepidemiology, 89: 572- 582 (1969). 14. KHAN, M. & MOSELY, W. Contrasting epidemiologic patterns of diarrhea and cholera in a semi-urban com- munity. Journal of the Pakistan Medical Association, 19: 380- 385 (1969). 15. OsEAsoHN, R. Clinical and bacteriological findings among families of cholera patients. Lancet, 1: 340- 342 (1966). 16. BART, K. ET AL. Seroepidemiologic studies during a simultaneous epidemic of infection with El Tor Ogawa and classical Inaba Vibrio cholerae. Journal of infec- tious diseases, 121 (suppl): S17 - S24 (1970). 404 J. M. HUGHES ET AL. 17. KHAN, M. & SHAHIDULLA, M. Pattern of intrafamilial spread of cholera. In: Proceedings of the 14th Joint Conference of the U.S. -Japan Cooperative Medical Science Program Cholera Panel Symposium on Cholera, Karatsu, 1978, Tokyo, Toho University, 1979, pp. 30- 34. 18. FELSENFELD, 0. Notes on food, beverages and fomites contaminated with Vibrio cholerae. Bulletin of the World Health Organization, 33: 725- 734 (1965). 19. LEVINE, R. ET AL. Cholera transmission near a cholera hospital. Lancet, 1: 84- 86 (1976). 20. BANG, F. ET AL. Ecology of respiratory virus trans- mission: A comparison of three communities in West Bengal. American journal of tropical medicine and hygiene, 24: 326- 346 (1975).